Latency improvement for file transfers over network connections

Abstract

A method of improving latency time of a data transfer between a sender and a receiver, receiving an odd number of data segments from among a plurality of data segments the receiver determines if it is waiting for a further data segment of less than full size. The determination is based on a total number of data segments received, an amount of information expected in the data transfer, and a maximum segment size. If the receiver is waiting for a further data segment of less than full size then sending a message is sent to the sender that triggers the transmission by the sender of said less than full size data segment.

Description

FIELD OF THE INVENTION[0001]The present invention relates generally to latency improvement over network connections where a network transport protocol implements the Nagle algorithm for the transfer of data. The invention relates more specifically to receiver-side control of deadlocks due to Nagle where packet segmentation results in an odd number of packets followed by a partial packet.BACKGROUND OF THE INVENTION[0002]The approaches described in this section could be pursued, but are not necessarily approaches that have been previously conceived or pursued. Therefore, unless otherwise indicated herein, the approaches described in this section are not prior art to the claims in this application and are not admitted to be prior art by inclusion in this section.[0003]Transport Control Protocol (TCP) as defined in IETF RFC 793 specifies a transport layer protocol for data networks. TCP generally consists of a set of rules defining how entities interact with each other. The OSI network ...

AI Technical Summary

Benefits of technology

[0025]Techniques are provided for modifying data processing performed by a TCP receiver to accept feedba

Problems solved by technology

Given the overhead associated with TCP, explicit acknowledgment of every segment, could generate significant extra traffic along the connection.

A problem arises according to the above TCP implementation.

Specifically, when a large upload data transfer is initiated over a HTTPS connection and the TCP connection has the Nagle algorithm implemented, a latency problem arises due to the segmentation logic of the TCP sender and the TCP receiver's delay acknowledgment logic.

High latency occurs in such a transfer due to the cumulative effect of several of the 200 ms delays, in some cases one delay per data record transmitted.

This problem is also found in HTTPS communications in which the sender's SSL or TSL layer generates large records.

Additionally, in many cases the TCP receiver does not have control over the TCP sender's logic.

Therefore, is it not practical in many cases to turn off Nagle at the sender or implement a TCP sender-based solution.

Images